Networking now determines how effectively memory reaches GPUs Introduction Memory has become a critical constraint in AI infrastructure. As models scale, solutions involve a mixture of experts, and context windows expand, the way memory is used must change. If we don’t change, GPUs stall waiting for data rather than computing.… Read More »Marvell Expands AI Memory and Storage Infrastructure with Bravera SC6
Alan Weckel had a great conversation with Sandeep Bharathi, President of the Data Center Group at Marvell, during their Industry Analyst Day. Sandeep provides an in-depth perspective on what Marvell has been up to in custom silicon, along with the momentum they see on XPU attach to alleviate bottlenecks in… Read More »Video Interview: Marvell’s Sandeep Bharathi Speaks with 650 Group
We attended the Marvell Industry Analyst Day in Santa Clara, California today. The company focused almost exclusively on the Data Center market opportunities the company is pursuing. Numerous spokespersons said that it thinks it has enough manufacturing capacity to hit its 2026 plans and it is working with customers and its suppliers… Read More »Marvell’s CPO Story Unfolds at its Industry Analyst Meeting
This week at NVIDIA’s GTC show in Washington, D.C., NVIDIA and Nokia announced that NVIDIA will invest $1 billion in Nokia to jointly pursue the goal of developing AI-RAN systems in the future. Based on the companies’ formal announcement and a call hosted by Nokia earlier this week, we understand that Nokia… Read More »NVIDIA Invests $1B In Nokia to Influence AI-RAN
Scale-Up (Ethernet, NVLink, PCIe, UALink), Scale-Out (Ethernet, InfiniBand), and Front-End (Ethernet) Contribute to $50B+ in Incremental Networking Opportunity This Decade AI’s extreme performance demands are driving divergence in networking architectures, featuring distinct protocols, technologies, and topologies across components. We expect companies to embrace multiple aspects of these challenges and develop… Read More »A Tale of Three Networks, AI Networking Infrastructure Surge Spans Multiple Protocols and Technologies
First Delivery of Silicon Indicates Technology Innovation Cycles are Accelerating The innovation and product cycles for Cloud infrastructure continue to push the limits of all technologies as Hyperscalers push AI to 1M XPU clusters and squeeze out efficiencies in order to scale power. The push towards AI will fundamentally change… Read More »Marvell 2nm to Usher in Next-Generation Platforms to Help Drive AI Workloads
As We Approach $100B/Hyperscaler in CAPEX, Supply Chain Will Continue to Adjust As Hyperscalers continue to push the envelope of what is technically feasible, we will see more announcements like the one earlier this month between Amazon and Marvell. The need to drive leading-edge technology at a more rapid pace… Read More »Marvell’s Five-Year Announcement with Amazon’s AWS Showcases Increased Partnerships Between Suppliers and Hyperscalers
Custom 5nm NIC Displays Impressive Performance Marvell and Meta announced the co-development of the FBNIC. The FBNIC is a custom 5nm NIC ASIC built to optimize the performance of its infrastructure for general purpose servers. Marvell contributed the board design to the OCP community. The announcement was significant for the… Read More »Marvell Showcases Meta FBNIC at OCP 2024
Taking Networking Power Consumption to Zero Won’t Allow You to Get More AI Into Your Data Centers With the emergence of AI/ML workloads, there is an industry-wide sense of urgency and panic about how we will power all these new servers. As such, there is tremendous pressure on all equipment… Read More »When Thinking About AI Power Consumption, Networking is the Least of Your Problems
Retimers Needed as Distances Shrink in PCIe Gen 5 and Gen 6 SpeedsThe server market remained on Gen 3 (8 Gbps/lane) for several server design cycles, leading the market to become complacent about how quickly server bandwidth could increase inside the server and at the NIC level. The market saw… Read More »PCIe Retimers Market to Surge Driven by AI/ML Computing
